Protein Biomarkers for in Vitro Testing of Embryotoxicity Karlfried Groebe,† Katrin Hayess,§ Martina Klemm-Manns,† Gerhard Schwall,† Woijciech Wozny,† Margino Steemans,# Annelieke K. Peters,# Chaturvedala Sastri,† Petra Jaeckel,† Werner Stegmann,† Helmut Zengerling,† Rainer Scho ¨ pf,† Slobodan Poznanovic,† ⊥ § § Tina C. Stummann, Andrea Seiler, Horst Spielmann, and Andre´ Schrattenholz*,† ProteoSys AG, Carl-Zeiss.-Str. 51, D-55129 Mainz, Germany, Federal Institute for Risk Assessment, Center for Alternative Methods to Animal Experiments - ZEBET Diedersdorfer Weg 1, 12277 Berlin, Germany, Johnson & Johnson PRD, a division of Janssen Pharmaceutical, 2340 Beerse, Belgium, European Centre for the Validation of Alternative Methods (ECVAM) (IHCP, JRC), Via Fermi, 121020 Ispra, Italy Received May 21, 2010
There are new challenges for hazard and risk assessment in the chemical industry with regard to REACH legislation in Europe and related activities in the U.S. and Japan, which require the development of novel in vitro models for the molecular characterization of drug- or chemical-related effects replacing conventional animal testing. In the frame of a European FP6 project on reproductive toxicology (www.reprotect.eu), we prepared protein samples from mouse embryonic stem cells differentiated into contracting cardiomyocytes according to the validated embryonic stem cell test (EST) protocol, which had been exposed to toxic substances selected by an expert committee from different in vivo categories of embryotoxicity. Lysates were used to carry out the following investigations: (i) identify optimal dose range conditions in the EST that are suitable for (ii) performing a differential quantitative proteomic study of underlying molecular pathways, (iii) define classes of substances with similar proteomic response patterns, (iv) relate these classes to the traditional in vivo categories of embryotoxicity with (v) the final goal to identify novel surrogate protein biomarker candidates for embryo toxicity. We found two distinct classes of toxic substances (Dinoseb, Ochratoxin-A, and Nitrofen vs β-aminoproprionitril, Metoclopramide, Doxylamine succinate, and D-penicillamine) with clear pathway-related differences in their proteomic patterns. Most notably, different responses to cluster 1 and cluster 2 substances were observed for Heat shock protein β-1, Ras-GTPase-activating protein SH3-domain binding protein, Ran binding protein 5, and Calreticulin, Dihydropyrimidinase-like 2 (Ulip2 protein). On the other hand, Heat shock protein 8 and Fscn1 protein were down-regulated by all compounds from both clusters. Keywords: Embryonic stem cell test • developmental toxicology • toxicity biomarker • Ras pathway • differential proteomics • REACH • chemicals testing
Introduction Safety-related guide lines of the OECD require in vivo tests for embryotoxicity. The special vulnerability of the developing human central nervous system has led to recommendations to test substances with a known neurotoxic or neuroteratogenic risk in the frame of the REACH program of the European Union and in related guide lines of the U.S. EPA (test guide line 870.6300) and the OECD (OECD draft guide line 426).1 However, end points in animal models often do not reliably and unambiguously reflect mechanisms in humans. Here we show that combining quantitative proteomic differences and functional data from appropriate in vitro models has the potential to provide surrogate biomarkers. The main * Corresponding author: Andre´ Schrattenholz, ProteoSys AG; phone, +496131-5019215; e-mail, andre.schrattenholz@proteosys@com. † ProteoSys AG. § Federal Institute for Risk Assessment. # Johnson & Johnson PRD. ⊥ European Centre for the Validation of Alternative Methods (ECVAM). 10.1021/pr100514e
2010 American Chemical Society
data set was generated by the European FP6 project on reproductive toxicology (www.reprotect.eu)2 using the embryonic stem cell test (EST), a model validated by an ECVAM study3 and a set of model substances assigned to four categories of embryotoxicity “strong”, “moderate”, “mild”, or “non-embryotoxic” based on in vivo data (substances were selected according to a peer-reviewed process by independent experts with industrial, academic or regulatory background4). Substancedependent cardiomyocyte protein extracts were subjected to systematic differential proteomic profiling. Dual radioisotope labeling of proteins provided the rigorous quantitative pattern control necessary to obtain statistical significance.5 This is the first time that a systematic proteomic differential of the molecular effects of a specific collection of embryotoxic substances has been performed in a validated in vitro model. Specifically, the objectives were: (1) use EST as a functional correlate of embryotoxicity in order to define optimal dose ranges of each substance for subsequent proteomic analyses, (2) perform a quantitative differential proteomic analysis of Journal of Proteome Research 2010, 9, 5727–5738 5727 Published on Web 10/04/2010
research articles these end points, (3) identify classes of similar response patterns, and (4) relate these classes to the traditional in vivo categories of embryotoxicity. The aim of this work is to identify biomarker candidates for novel and fast in vitro strategies for safety tests for developmental toxicity, with the potential to substantially reduce animal experiments according to the “3R” concept (Reduce/ Refine/Replace).6 The Ras-related and structural proteins found here need to be further characterized with regard to posttranslational modifications. We discuss necessary steps to carry this early discovery phase towards validation and translation to hig- throughput in vitro methods of pathway-based screening, including: (1) further characterization of the molecular detail underlying the different clusters of embryotoxic substances, in particular post-translational modifications; (2) expansion to other organ-specific in vitro models; (3) validation, establishment of predictive values, and integration of independent methods; (4) establishment of novel test tools for assessing the severity of embryotoxic substance effects in vitro (e.g. ELISA).
Experimental Section Protein lysates were generated from a series of experiments applying the EST in two independent laboratories. In accordance with the standardized EST protocol, contracting cardiomyocytes differentiated from mouse embryonic stem cells were exposed to sets of model substances with known embryotoxic potency.3,7 Briefly, 750 D3 cells in 20 µL of supplementedDulbecco’s minimal essential medium (DMEM; without mLIF to allow differentiation) were placed as hanging drops on the lid of a Petri dish filled with phosphate buffered saline (PBS, Sigma, St. Louis, MO) and then incubated for 3 days at 37 °C under 5% CO2 and 95% humidity in the presence of test chemicals at various concentrations. During this period, the cells formed aggregates referred to as embryoid bodies (EBs). After 3 days of “hanging drop” culture, the EBs were transferred to bacterial Petri dishes (Greiner, Frickenhausen, Germany) containing the appropriate concentration of the test chemical for another 2 days. Bacterial Petri dishes were used to avoid adherence and outgrowth of the EBs during this period in culture. On day 5, EBs were plated separately into the wells of a 24-multiwell tissue culture plate (containing the appropriate concentration of test chemical) to allow adherence and outgrowth of the EBs and development of spontaneously beating cardiac muscle cells. Differentiation was determined by microscopic inspection of EB outgrowths at day 10 of differentiation (validated end point). The number of wells containing contracting cardiomyocytes was determined for each plate and compared to the number of wells containing contracting cardiomyocytes in a solvent control plate. Experiments were accepted when solvent control plates showed beating cardiomyocytes in 21 out of 24 wells. The analysis was performed using phase-contrast microscopy. The end point of the EST is described by the “inhibition of differentiation” (ID50) which is expressed as the concentration of the test chemical that reduces the number of wells that contain contracting cardiomyocytes by 50% (calculated from the concentration response curve). The analysis of dose-response curves was done in the R statistical computing environment (R Development Core Team, 2009) using the R package drc.8 In a similar fashion as ID50, IC50D3 and IC503T3 are defined as 50% loss of ES cell viability or of 3T3 cell viability, 5728
Journal of Proteome Research • Vol. 9, No. 11, 2010
Groebe et al. respectively. To determine cytotoxic effects on ES and 3T3 cells, MTT-assays were performed in the absence of mLIF as described previously.9 Briefly, 500 cells were seeded into each well of a 96-well microtiter plate and grown in the presence of a concentration range of the test chemical. A negative control containing solvent diluted in medium was also included. After 10 days in culture and two changes of medium (containing the appropriate concentration of test chemical) at days 3 and 5, the viability of the cells was determined using an MTT test, which is based on the capacity of mitochondrial dehydrogenase enzymes in living cells to convert the yellow substrate 3-(4,5dimethylthiazol-2yl)-2,5-diphenyl tetrazolium bromide (MTT) into a dark blue formazan product, which was detected quantitatively using a microplate ELISA reader. Cytotoxicity is expressed as the concentration decreasing viability to 50% of the control level (IC503T3 and IC50D3; determined from a concentration-response curve). On the basis of the EST results, substance concentrations for the differential proteomic analyses were chosen from the series of concentrations applied in the EST in such that they were as close as feasible to the ID50 value (rationale: optimal response amplitude, but sufficient surviving cells for protein lysates). ID50 was obtained by interpolating the concentration response curve. Fifty embryoid bodies from mESC differentiated into cardiomyocytes were plated into a 100 mm cell culture treated dish on day 5. On day 7 of differentiation, substance-treated cells as well as solvent-treated cells were scraped in lysis buffer (2% SDS; 100 mM Tris, pH 8.8; 1× Complete Proteinase Inhibitor Mix (Roche, Mannheim), and 1 mM activated orthovanadate) and heated to 95 °C for 5 min. After addition of 5 mM MgCl2 and benzonase (250 U/mL, Merck, Jena), DNA was sheared using a G27 injection needle. The lysates were centrifuged for 20 min at 20 000g and the supernatants were subjected to differential proteome analysis. Substance conditions tested: D-Penicillamine (800 µg/mL), Methylazoxymethanol (50 µg/mL), Nitrofen (22 µg/mL), Ochratoxin-A (10 µg/mL), Lovastatin (5 µg/mL), Warfarin (200 µg/ mL), Dinoseb (18 µg/mL), Furosemide (667 µg/mL), β-aminoproprionitril(667µg/mL),Doxylamine(80µg/mL),Metoclopramide (67 µg/mL), Pravastatin (50 µg/mL). Substances had to be applied in different solvents (EtOH for Dinoseb and OchratoxinA; DMSO for Furosemide, Nitrofen, Lovastatin, and Warfarin; PBS for all other substances) and solvent controls were included in the analysis. The differential proteomic technology for the quantitative and statistical analysis of protein biomarkers by dual radioisotope labeling, 2D gel electrophoresis, and the subsequent analysis by MALDI-TOF mass spectrometry has been described in detail recently.5,10 Briefly, minute amounts of two different samples are labeled by two isotopes of iodine (125I and 131I), analyzed together in single separations (high resolution 2DPAGE), and subsequently analyzed by protein spot detection and independent quantifications of both label abundances. For statistical purposes and correction of label induced bias, for each sample pair, two inversely labeled gels were prepared in replicate (N ) 4 for each experimental condition). The large dynamic range and quantitative accuracy of radioactive imaging techniques allow for rigorous statistical analysis of differences between conditions and the use of pooling schemes.10 In this work, single samples were directly profiled against a reference solvent pool which consisted of a fixed preparation of differentiated stem cells kept in plain medium. Hence, in each gel, a mixture of one treated sample and the reference
Embryotoxicity Protein Biomarkers pool were directly compared to each other by calculating the in-gel abundance ratios for each protein spot. Substance effects were evaluated by quantifying the changes in protein abundances between samples treated with the respective substance and the corresponding solute control, that is, by forming the suitable ratios.10 Hence, the samples that were effectively compared to each other had been harvested at exactly the same stages in their differentiation process and had undergone identical treatments, except that a substance was applied to one of them and the corresponding solute control to the other. Whenever a protein abundance was significantly higher in substance treated samples as compared to corresponding solute controls, this was denoted an “up-regulation” under treatment. When protein abundance under treatment was lowered, this was noted a “down-regulation”. Analysis of the respective gel images was performed to detect protein spots using the Pic/Greg software package by the Fraunhofer Gesellschaft in Sankt Augustin. A spot is selected for further analysis if t test indicates different abundances for the compared conditions (p-value e0.01), and the abundance ratio of the compared spot intensities is >1.5 or
